关键词: SK算法, 凸包, 支持向量机, 几何方法, 数据挖掘

Abstract: The geometric approach of the Support Vector Machine（SVM） is a kind of geometric way to find the solution to the problem of the SVM algorithm.Based on its geometric characters，the SK（Schlesinger-Kozinec） algorithm is studied intuitively. It briefly sums up the two convex hulls，based on their relative positions，into five categories，and makes sure their optimizing position got in each computing is mostly at the hull vertices or boundary，it can get to the boundary（the optimization place of the computing） at the first computation.The manual single-step simulation results show that the projection is not always successful for such kind of algorithms in many cases，though it can’t affect the computing result，but can weaken the algorithm efficiency.Based on the analysis，it demonstrates two improving ways for the soft SK algorithm（Backward-SK and Forward-SK methods），and makes some simulation for comparing.The simulation results show that the improved method computing results are almost same as the SK and soft SK ones，but the computing process of improved one is more intuitive.

Key words: SK algorithm, convex hull, support vector machine, geometric approach, data mining